cryogenic is a package for freezing a function call, modifying its
arguments and then evaluating later.
This package is written for a particular use case I have in mind, and is not trying to be the comprehensive answer to call manipulation in R.
- Both the calling environment and evaluation environment are under my
control.
- This package is not targetted at executing captured calls based on possibly hostile user input.
- Need an R package for this as I need to use this code across multiple packages
- Need eager evaluation of arguments when call is captured
- Not concerned with lazy evaluation of the call arguments.
- Need to be able to update any arguments
- Need to be able to update an argument to have a value of
NULL
- Need to be able to update an argument to have a value of
- Need to be able to set arguments to a default value if they aren’t already present
- Nearly everything is in lists
- Need to be able to consistently add arbitrary
metainformation as an attribute on the call
capture_call()to capture a call without evaluating itmodify_call()to modify call argumentsevaluate_call()to evaluate a call - and optionally modify the call arguments just prior to the evaluation.
You can install from GitHub with:
# install.package('remotes')
remotes::install_github('coolbutuseless/cryogenic')- capture a call
- modify the call
- evaluate the call
library(cryogenic)
cc <- capture_call(runif(n = 10, min=-3))
cc#> runif(n = 10, min = -3)
cc <- modify_call(cc, defaults = list(min=0, max=5), update = list(n = 5))
cc#> runif(n = 5, min = -3, max = 5)
evaluate_call(cc)#> [1] -0.8759307 -0.0230088 1.5828269 4.2656623 -1.3865446
- capture a call
- simultaneously update a call and evalute it
library(cryogenic)
cc <- capture_call(runif(n = 10, min=-3))
cc#> runif(n = 10, min = -3)
evaluate_call(cc, defaults = list(min=0, max=5), update = list(n = 15))#> [1] 4.18711748 4.55740215 2.28638234 2.03291235 -2.50570984 -1.35220340
#> [7] -1.58754598 2.49618277 0.07282975 3.15873136 0.98159394 2.74094807
#> [13] 4.93524876 0.04028144 3.21956177
library(cryogenic)
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Capture a call as-is.
#
# * the function does not have to exist in the capturing environment
# * By default, all arguments are eagerly evaluated in the environment in which
# the call was captured
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc <- capture_call(farnarkle(6, x = 2, y = 3, z = 2+9, error = TRUE))
cc#> farnarkle(6, x = 2, y = 3, z = 11, error = TRUE)
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update the arguments to the call
#
# * 'defaults' are only added to the arguments if the named argument does
# not already exist there.
# * 'update' values are added to the arguments unconditionally
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc <- modify_call(
cc,
defaults = list(x = NULL, y = 7),
update = list(z = NULL, na.rm = TRUE),
delete = c('error')
)
cc#> farnarkle(6, x = 2, y = 3, z = NULL, na.rm = TRUE)
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# 'farnarkle' is not yet defined, so evaluation should fail
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
evaluate_call(cc)#> Error in farnarkle(6, x = 2, y = 3, z = NULL, na.rm = TRUE): could not find function "farnarkle"
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# define farnarkle and it should work
#~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
farnarkle <- function(...) {print("Hello #Rstats")}
evaluate_call(cc)#> [1] "Hello #Rstats"
rlangoffers a lot of tools for call modification, but wasn’t quite a match for how I’m working.pryr::modify_call()is similar to what is done in this package, butcryogenicoffers a few more features I need.
- R Core for developing and maintaining the language.
- CRAN maintainers, for patiently shepherding packages onto CRAN and maintaining the repository